Foundations of Robotics
Seminar, November 17, 2009
Time
and Place | Seminar Abstract
A Proof for Probabilistically Complete Planning with End-Effector Pose
Constraints
Dmitry Berenson
RI
CMU
NSH 1507
Talk 4:30 pm
We present a proof for the probabilistic completeness of the Constrained
BiDirectional RRT (CBiRRT) when planning with constraints on end-effector
pose. Pose constraints can induce lower-dimensional constraint manifolds
in the configuration space of the robot, making rejection-sampling
techniques infeasible. The CBiRRT overcomes this problem by using random
sampling coupled with projection methods to move configuration space
samples onto the constraint manifold. Until now it was not known whether
this sampling scheme produced adequate coverage of the manifold to
guarantee probabilistic completeness. The proof presented in this talk
guarantees probabilistic completeness for the CBiRRT, as well as other
sampling-based planners, given an appropriate projection operator. This
proof is valid for any end-effector pose constraint, regardless of the
dimensionality of the manifold induced by the constraint.
The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.